Thread: [libposix-development] [PATCH 2/2] x86_32: use 'int 0x80' for system calls to improve protability
Status: Pre-Alpha
Brought to you by:
hdante
Menu
▾
▴
libposix-development
|
From: Kirill A. S. <ki...@sh...> - 2009-06-26 14:44:20
|
Not every kernel provide __kernel_vsyscall. It's possible to add
configuration option later.
Signed-off-by: Kirill A. Shutemov <ki...@sh...>
---
runtime/linux/x86_32/architecture_init.c | 22 +---------------------
system/linux/x86_32/system_calls.S | 17 ++---------------
2 files changed, 3 insertions(+), 36 deletions(-)
diff --git a/runtime/linux/x86_32/architecture_init.c
b/runtime/linux/x86_32/architecture_init.c
index f080dc4..f3fb6e6 100644
--- a/runtime/linux/x86_32/architecture_init.c
+++ b/runtime/linux/x86_32/architecture_init.c
@@ -27,28 +27,8 @@ POSSIBILITY OF SUCH DAMAGE.
#include <stddef.h>
#include "runtime.h"
-void (*kernel_vsyscall)();
-
-/* AT_SYSINFO entry has the vsyscall address */
-#define AT_SYSINFO 32
-
-void set_kernel_vsyscall(const auxv_t *auxv)
-{
- if (auxv != NULL) {
- const auxv_t *auxv_walk;
- for (auxv_walk = auxv; auxv_walk->a_type != 0; auxv_walk++) {
- if (auxv_walk->a_type == AT_SYSINFO) {
- kernel_vsyscall = auxv_walk->a_un.a_fcn;
- return;
-
- }
- }
- // ERROR: AT_SYSINFO entry not found
- }
-}
-
void architecture_init(int argc, const char **argv, const char **envp,
const auxv_t *auxv)
{
- set_kernel_vsyscall(auxv);
+ /* Do nothing */
}
diff --git a/system/linux/x86_32/system_calls.S
b/system/linux/x86_32/system_calls.S
index 8e1d6f9..f11f8ca 100644
--- a/system/linux/x86_32/system_calls.S
+++ b/system/linux/x86_32/system_calls.S
@@ -36,8 +36,6 @@ POSSIBILITY OF SUCH DAMAGE.
Paramenter sequence is: %ebx, %ecx, %edx, %esi, %edi and %ebp
See, for example:
http://www.linuxjournal.com/article/4048
-
- TODO: remove calculation for kernel_vsyscall (use a fixed address)
*/
@@ -45,12 +43,9 @@ syscall1:
pushl %ebp
movl %esp, %ebp
pushl %ebx
- call get_abs_address
- addl $_GLOBAL_OFFSET_TABLE_, %ebx
- movl kernel_vsyscall@GOT(%ebx), %ecx
movl 8(%ebp), %eax
movl 12(%ebp), %ebx
- call *(%ecx)
+ int $0x80
popl %ebx
popl %ebp
ret
@@ -60,20 +55,12 @@ syscall3:
movl %esp, %ebp
pushl %edi
pushl %ebx
- call get_abs_address
- addl $_GLOBAL_OFFSET_TABLE_, %ebx
- movl kernel_vsyscall@GOT(%ebx), %edi
movl 8(%ebp), %eax
movl 12(%ebp), %ebx
movl 16(%ebp), %ecx
movl 20(%ebp), %edx
- call *(%edi)
+ int $0x80
popl %ebx
popl %edi
popl %ebp
ret
-
-get_abs_address:
- mov (%esp), %ebx
- ret
-
--
1.6.3.3
|
|
From: Henrique A. <hd...@gm...> - 2009-06-26 17:21:19
|
Thanks for the patch. However, do you need those for linux 2.4 ? If
you do, that would be a variant from the default implementation. I'll
explain variants when I define platform sets, but the point is that
the amount of variants should be kept to a minimum possible. For linux
2.6, vsyscall can always be set and is considered to be the default
way to do system calls (because it's faster in modern processors), so
a second implementation would be redundant. Also, I wasn't considering
supporting linux 2.4 in x86. If someone is willing to switch to
libposix to use a newer standard, then probably the person is also
willing to update to linux 2.6 ;-)
So, if you DEFINITELY need x86 linux 2.4 support in libposix, please
resend the patch as a variant: create a subdirectory called "linux24"
in system/linux/x86_32 and runtime/linux/x86_32 and create a separate
cmake file to support the variant by redefining the necessary support
files.
2009/6/26 Kirill A. Shutemov <ki...@sh...>:
> Not every kernel provide __kernel_vsyscall. It's possible to add
> configuration option later.
>
> Signed-off-by: Kirill A. Shutemov <ki...@sh...>
> ---
> runtime/linux/x86_32/architecture_init.c | 22 +---------------------
> system/linux/x86_32/system_calls.S | 17 ++---------------
> 2 files changed, 3 insertions(+), 36 deletions(-)
>
> diff --git a/runtime/linux/x86_32/architecture_init.c
> b/runtime/linux/x86_32/architecture_init.c
> index f080dc4..f3fb6e6 100644
> --- a/runtime/linux/x86_32/architecture_init.c
> +++ b/runtime/linux/x86_32/architecture_init.c
> @@ -27,28 +27,8 @@ POSSIBILITY OF SUCH DAMAGE.
> #include <stddef.h>
> #include "runtime.h"
>
> -void (*kernel_vsyscall)();
> -
> -/* AT_SYSINFO entry has the vsyscall address */
> -#define AT_SYSINFO 32
> -
> -void set_kernel_vsyscall(const auxv_t *auxv)
> -{
> - if (auxv != NULL) {
> - const auxv_t *auxv_walk;
> - for (auxv_walk = auxv; auxv_walk->a_type != 0; auxv_walk++) {
> - if (auxv_walk->a_type == AT_SYSINFO) {
> - kernel_vsyscall = auxv_walk->a_un.a_fcn;
> - return;
> -
> - }
> - }
> - // ERROR: AT_SYSINFO entry not found
> - }
> -}
> -
> void architecture_init(int argc, const char **argv, const char **envp,
> const auxv_t *auxv)
> {
> - set_kernel_vsyscall(auxv);
> + /* Do nothing */
> }
> diff --git a/system/linux/x86_32/system_calls.S
> b/system/linux/x86_32/system_calls.S
> index 8e1d6f9..f11f8ca 100644
> --- a/system/linux/x86_32/system_calls.S
> +++ b/system/linux/x86_32/system_calls.S
> @@ -36,8 +36,6 @@ POSSIBILITY OF SUCH DAMAGE.
> Paramenter sequence is: %ebx, %ecx, %edx, %esi, %edi and %ebp
> See, for example:
> http://www.linuxjournal.com/article/4048
> -
> - TODO: remove calculation for kernel_vsyscall (use a fixed address)
> */
>
>
> @@ -45,12 +43,9 @@ syscall1:
> pushl %ebp
> movl %esp, %ebp
> pushl %ebx
> - call get_abs_address
> - addl $_GLOBAL_OFFSET_TABLE_, %ebx
> - movl kernel_vsyscall@GOT(%ebx), %ecx
> movl 8(%ebp), %eax
> movl 12(%ebp), %ebx
> - call *(%ecx)
> + int $0x80
> popl %ebx
> popl %ebp
> ret
> @@ -60,20 +55,12 @@ syscall3:
> movl %esp, %ebp
> pushl %edi
> pushl %ebx
> - call get_abs_address
> - addl $_GLOBAL_OFFSET_TABLE_, %ebx
> - movl kernel_vsyscall@GOT(%ebx), %edi
> movl 8(%ebp), %eax
> movl 12(%ebp), %ebx
> movl 16(%ebp), %ecx
> movl 20(%ebp), %edx
> - call *(%edi)
> + int $0x80
> popl %ebx
> popl %edi
> popl %ebp
> ret
> -
> -get_abs_address:
> - mov (%esp), %ebx
> - ret
> -
> --
> 1.6.3.3
>
> ------------------------------------------------------------------------------
> _______________________________________________
> Libposix-development mailing list
> Lib...@li...
> https://lists.sourceforge.net/lists/listinfo/libposix-development
>
--
Henrique Dante de Almeida
hd...@gm...
|
|
From: Kirill A. S. <ki...@sh...> - 2009-06-26 17:30:07
|
On Fri, Jun 26, 2009 at 8:20 PM, Henrique Almeida<hd...@gm...> wrote: > Thanks for the patch. However, do you need those for linux 2.4 ? No, I use 2.6.30. Config here: http://git.altlinux.org/people/lakostis/packages/?p=kernel-image-2.6.30.git;a=blob;f=config-i586;h=9635402ce20ee00645daeb3ffd1f236fde01e5cd;hb=fc05c2cdc66d7d64a62c5f07df43e25437a5902e I guess it because CONFIG_COMPAT_VDSO=n. |
|
From: Henrique A. <hd...@gm...> - 2009-06-26 17:33:37
|
Ok, is the config default from your distro ? Or can you turn the vdso on ? 2009/6/26 Kirill A. Shutemov <ki...@sh...>: > On Fri, Jun 26, 2009 at 8:20 PM, Henrique Almeida<hd...@gm...> wrote: >> Thanks for the patch. However, do you need those for linux 2.4 ? > > No, I use 2.6.30. Config here: > > http://git.altlinux.org/people/lakostis/packages/?p=kernel-image-2.6.30.git;a=blob;f=config-i586;h=9635402ce20ee00645daeb3ffd1f236fde01e5cd;hb=fc05c2cdc66d7d64a62c5f07df43e25437a5902e > > I guess it because CONFIG_COMPAT_VDSO=n. > > ------------------------------------------------------------------------------ > _______________________________________________ > Libposix-development mailing list > Lib...@li... > https://lists.sourceforge.net/lists/listinfo/libposix-development > -- Henrique Dante de Almeida hd...@gm... |
|
From: Kirill A. S. <ki...@sh...> - 2009-06-26 17:37:54
|
On Fri, Jun 26, 2009 at 8:32 PM, Henrique Almeida<hd...@gm...> wrote: > Ok, is the config default from your distro ? Or can you turn the vdso on ? It seems only for compatibility with old glibc: config COMPAT_VDSO def_bool y prompt "Compat VDSO support" depends on X86_32 || IA32_EMULATION ---help--- Map the 32-bit VDSO to the predictable old-style address too. ---help--- Say N here if you are running a sufficiently recent glibc version (2.3.3 or later), to remove the high-mapped VDSO mapping and to exclusively use the randomized VDSO. If unsure, say Y. I don't this that it's right way. |
|
From: Henrique A. <hd...@gm...> - 2009-06-26 17:35:50
|
Wait, my CONFIG_COMPAT_VDSO is not set either. 2009/6/26 Henrique Almeida <hd...@gm...>: > Ok, is the config default from your distro ? Or can you turn the vdso on ? > > 2009/6/26 Kirill A. Shutemov <ki...@sh...>: >> On Fri, Jun 26, 2009 at 8:20 PM, Henrique Almeida<hd...@gm...> wrote: >>> Thanks for the patch. However, do you need those for linux 2.4 ? >> >> No, I use 2.6.30. Config here: >> >> http://git.altlinux.org/people/lakostis/packages/?p=kernel-image-2.6.30.git;a=blob;f=config-i586;h=9635402ce20ee00645daeb3ffd1f236fde01e5cd;hb=fc05c2cdc66d7d64a62c5f07df43e25437a5902e >> >> I guess it because CONFIG_COMPAT_VDSO=n. >> >> ------------------------------------------------------------------------------ >> _______________________________________________ >> Libposix-development mailing list >> Lib...@li... >> https://lists.sourceforge.net/lists/listinfo/libposix-development >> > > > > -- > Henrique Dante de Almeida > hd...@gm... > -- Henrique Dante de Almeida hd...@gm... |
|
From: Henrique A. <hd...@gm...> - 2009-06-26 17:37:05
|
Can you check /proc/sys/vm/vdso_enabled ? 2009/6/26 Henrique Almeida <hd...@gm...>: > Wait, my CONFIG_COMPAT_VDSO is not set either. > > 2009/6/26 Henrique Almeida <hd...@gm...>: >> Ok, is the config default from your distro ? Or can you turn the vdso on ? >> >> 2009/6/26 Kirill A. Shutemov <ki...@sh...>: >>> On Fri, Jun 26, 2009 at 8:20 PM, Henrique Almeida<hd...@gm...> wrote: >>>> Thanks for the patch. However, do you need those for linux 2.4 ? >>> >>> No, I use 2.6.30. Config here: >>> >>> http://git.altlinux.org/people/lakostis/packages/?p=kernel-image-2.6.30.git;a=blob;f=config-i586;h=9635402ce20ee00645daeb3ffd1f236fde01e5cd;hb=fc05c2cdc66d7d64a62c5f07df43e25437a5902e >>> >>> I guess it because CONFIG_COMPAT_VDSO=n. >>> >>> ------------------------------------------------------------------------------ >>> _______________________________________________ >>> Libposix-development mailing list >>> Lib...@li... >>> https://lists.sourceforge.net/lists/listinfo/libposix-development >>> >> >> >> >> -- >> Henrique Dante de Almeida >> hd...@gm... >> > > > > -- > Henrique Dante de Almeida > hd...@gm... > -- Henrique Dante de Almeida hd...@gm... |
|
From: Kirill A. S. <ki...@sh...> - 2009-06-26 19:53:34
|
On Fri, Jun 26, 2009 at 8:36 PM, Henrique Almeida<hd...@gm...> wrote: > Can you check /proc/sys/vm/vdso_enabled ? I can't reproduce the problem on my home machine. I'll try it on Monday at work. It looks like local misconfiguration. |
|
From: Henrique A. <hd...@gm...> - 2009-06-26 20:56:44
|
2009/6/26 Kirill A. Shutemov <ki...@sh...>: > On Fri, Jun 26, 2009 at 8:36 PM, Henrique Almeida<hd...@gm...> wrote: >> Can you check /proc/sys/vm/vdso_enabled ? > > I can't reproduce the problem on my home machine. I'll try it on Monday at work. > It looks like local misconfiguration. Ok. Can you send patch 1/2 again ? > > ------------------------------------------------------------------------------ > _______________________________________________ > Libposix-development mailing list > Lib...@li... > https://lists.sourceforge.net/lists/listinfo/libposix-development > -- Henrique Dante de Almeida hd...@gm... |
|
From: Henrique A. <hd...@gm...> - 2009-07-01 02:17:41
|
Were you able to solve this issue ? 2009/6/26 Kirill A. Shutemov <ki...@sh...>: > On Fri, Jun 26, 2009 at 8:36 PM, Henrique Almeida<hd...@gm...> wrote: >> Can you check /proc/sys/vm/vdso_enabled ? > > I can't reproduce the problem on my home machine. I'll try it on Monday at work. > It looks like local misconfiguration. > > ------------------------------------------------------------------------------ > _______________________________________________ > Libposix-development mailing list > Lib...@li... > https://lists.sourceforge.net/lists/listinfo/libposix-development > -- Henrique Dante de Almeida hd...@gm... |
|
From: Kirill A. S. <ki...@sh...> - 2009-07-01 07:11:13
|
On Wed, Jul 1, 2009 at 5:10 AM, Henrique Almeida<hd...@gm...> wrote: > Were you able to solve this issue ? Yep. It's local problem. |
×
Want the latest updates on software, tech news, and AI?
Get latest updates about software, tech news, and AI from SourceForge directly in your inbox once a month.
Thanks for helping keep SourceForge clean.
X